Application of ontology to knowledge management of sucker rod pumping system fault diagnosis

Journal of Automation and Control Engineering, Vol. 1, No. 2, June 2013<br /> <br /> Application of Ontology to Knowledge<br /> Management of Sucker-rod Pumping System<br /> Fault Diagnosis<br /> Peng Cheng<br /> Tsinghua University, Beijing, PRC<br /> Email: pengchengthu@foxmail.com<br /> <br /> Yan Jian<br /> Xi’an Shiyou University, Shanxi Xi’an, PRC<br /> Email: 704874024qq.com<br /> <br /> Abstract—Sucker-rod pumping system operating status<br /> fault diagnosis technology is closely related to the energy<br /> consumption, production per well and the cost of crude oil<br /> exploration. But the knowledge depiction of the expert<br /> diagnostic system is mainly based on the production rules,<br /> neural networks or object-oriented. As ontology has<br /> advantages in hierarchy and semantic relationships<br /> expressing, it is can be used in the knowledge management<br /> in the expert diagnostic system of the fault diagnosis<br /> technology. Then an ontology model for sucker-rod<br /> pumping system running state fault diagnosis system is also<br /> proposed in this paper. Basing on ontology, the knowledge<br /> management system framework is built up, which includes<br /> data layer, logic layer and application layer. The<br /> comprehensive diagnosis ontology database, including the<br /> knowledge domains of fault, collection and structure, is<br /> established. And the relationships among the ontology are<br /> described. <br /> <br /> this technology appeared after the emerge of expert<br /> systems [13], [14], [22], [23], neural networks [15] and<br /> pattern recognition technology [16]-[21].<br /> The expert diagnosis system developed by H. J. Derk<br /> [13] has three main auxiliary programs: API, DHOLE<br /> and VIEW. By aid of the auxiliary programs, some tasks<br /> can be done: carry out the designs for conventional<br /> pumping device to get the rated operating parameters<br /> which are used as the reference data, calculate the<br /> hydrodynamic forces and inlet pressure of the pump<br /> basing on the transformation from the ground indicator<br /> diagram into downhole one, and so on. It also can help<br /> users to find out one certain fault group similar to the<br /> downhole indicator diagrams from all typical indicator<br /> diagrams in various kinds of fault group to check all<br /> indicator diagrams belong to this fault group. Then users<br /> can determine the fault category and get expert’s explain<br /> and corresponding countermeasure.<br /> The expert system developed by W. L. Foley et al. [14]<br /> is an auxiliary function module of the surface and<br /> downhole analysis program (SADA) which is a computer<br /> diagnostic technology. It includes a statistics-type model<br /> diagrams and a group of production rules featuring the<br /> form of "If-then". They are built up according to the<br /> expert experience and knowledge. The database has over<br /> 100 types of model diagrams corresponding to different<br /> typical failures of the pumping system, as well as more<br /> than 60 items of If-then rules of thumb, such as "If the<br /> downhole indicator diagram shows that the liquid level<br /> impact and pump inlet pressure is higher than the casing<br /> pressure, then it is indicated that the density of the liquid<br /> has changed or pump diameter is inappropriate ".<br /> The pumping system fault recognition software<br /> developed by R. R. Dickinson et al. [16] established a<br /> standard indicator diagram library having about 30 kinds<br /> of corresponding typical pumping system faults. It is<br /> based on the indicator diagrams features: boundary chain<br /> code, boundary curve Fourier descriptors and boundary<br /> gray matrix. Incorporating pattern recognition techniques<br /> and the criterion of minimum Euclidean distance, the user<br /> <br /> Index Terms—Ontology, pumping system, fault, knowledge<br /> management<br /> <br /> I.<br /> <br /> INTRODUCTION<br /> <br /> Sucker-rod pumping system (short of “pumping<br /> system”) running state fault diagnosis technology has<br /> been one of the important issues in oil production<br /> engineering. This technology can affect the exploitation<br /> cost, energy consumption and the production per well.<br /> After 80 years of studies and practices, it has been greatly<br /> developed, especially, with the emerge of computer<br /> diagnostic technology, has entered into a new phase, i.e.<br /> it has developed from the qualitative analysis to the<br /> quantitative analysis [1], at the same time, it has stepped<br /> into the intelligent diagnostic phase [2]. There are four<br /> develop stages in pumping system fault diagnosis<br /> technology: ground indicator diagram analysis [3]-[5],<br /> downhole indicator diagram diagnosis method [6],<br /> computer diagnostics [7-12] and artificial intelligence<br /> diagnostic method [13-23]. The practical applications of<br /> Manuscript received September 25, 2012; revised December 24,<br /> 2012.<br /> <br /> ©2013 Engineering and Technology Publishing<br /> doi: 10.12720/joace.1.2.126-131<br /> <br /> 126<br /> <br /> Journal of Automation and Control Engineering, Vol. 1, No. 2, June 2013<br /> <br /> can judge the operation status and diagnose, then give a<br /> proper countermeasure.<br /> Basing on the study of R. R. Dickinson et al., G. A. Yu<br /> et al. [17] established a standard indicator diagram library<br /> with 17 categories of typical pumping system faults. The<br /> classification and identification to the pumping system<br /> operation status by aid of the criterion of minimum<br /> Euclidean distance and absolute difference distance, the<br /> correct recognition rate is greatly improved. After further<br /> improvements, through extraction of the statistical nature<br /> of the indicator diagram boundary grayscale matrix [18]<br /> and incorporating the sample average self-learning<br /> function and Fisher linear classifier [19], the recognition<br /> speed and accuracy was further improved. In present, this<br /> technology is widely used in oilfields of China.<br /> Above mentioned diagnosis technologies and software<br /> are efficacious for single pumping system fault and it has<br /> verified an exciting economical benefit in increase crude<br /> oil production, reduce energy consumption and save<br /> workover cost. However, as to these techniques including<br /> expert system, neural network or pattern recognition, it is<br /> hard to carry out the knowledge and information<br /> acquisition, sensitivity and uncertainty reasoning, selflearning, and so on[24]. In other words, these techniques<br /> and software are immune to multi-factor fault. So it is<br /> necessary to develop an advanced diagnosis technology<br /> and software, which can be competent for multi-factor<br /> fault appeared in some production wells.<br /> II.<br /> <br /> ONTOLOGY AND ITS APPLICATION<br /> <br /> Ontology is one of the newly introduced fault<br /> diagnosis techniques in recent years. The conception of<br /> ontology originates from the philosophical conception<br /> which used to explain the nature of existence and<br /> relationship of objective things in ancient Greek<br /> philosophy [25] and [26]. On 17th century, Germany<br /> scholar first used ontology and interpreted it as a<br /> synonym of metaphysics. In accordance with the view of<br /> these scholars, ontology discourses various abstract and<br /> completely universal philosophical categories. Then from<br /> the abstract metaphysics, the duality, entity and causal<br /> phenomenon are derived. On a view of philosophy, it is<br /> used in the following two ways: "Yes" or "Existence". So,<br /> ontology is a innate principles system rely on the<br /> deducing from conception to conception [27]. It intends<br /> to uniformly conceptualize a certain definitional<br /> knowledge, then to look for the nature of all things from<br /> the natural internal and the relations among objects, and<br /> to build up the ontology of the objective world while<br /> braking away from the influence coming from human. In<br /> latest decade, the ontology is given a new definition and<br /> have gone far beyond the scope of philosophy. It is<br /> widely used in the fields of knowledge management,<br /> knowledge engineering and artificial intelligence. In<br /> modern information technology, ontology is defined as<br /> “conceptualization standardized explanation” [28]. These<br /> experts and professors, coming from the Knowledge<br /> Systems Laboratory of the Department of Computer<br /> Science in Stanford University, started a research plan<br /> named as “How Things Work” from the beginning of<br /> 127<br /> <br /> 1990s. The main purpose is to study the shared and<br /> reusable knowledge database which is for science and<br /> engineering and ontology-based. It has greatly promoted<br /> the researches of the ontology in knowledge engineering.<br /> Ontology-based knowledge representation includes<br /> five modeling primitives: conception, attribute, relation,<br /> axiom and instance. It combines the advantages of the<br /> traditional method of knowledge representation and has<br /> rich knowledge representation languages. This can well<br /> realize the sharing and reuse of knowledge to make up for<br /> the defects of the traditional knowledge representation.<br /> According to the characteristics of the domain knowledge,<br /> select a most appropriate method to descript knowledge,<br /> and in terms of the complexity of this field, combine<br /> several techniques to form a powerful system to<br /> efficiently solve the intelligent problems. Therefore, there<br /> are more scholars choose ontology-based knowledge<br /> representation to descript the knowledge in the domain of<br /> mechanical fault diagnosis [29]-[34].<br /> Y. Kitamura et al. [29] built up the ontology for steam<br /> engine fault diagnosis. The steam engine fault diagnosis<br /> ontology modeling has been investigated, including fault<br /> classification, fault representation. The aim is to develop<br /> an interactive system which can diagnose according to<br /> the users’ failure symptom description.<br /> A. Bernaras et al. [30] introduced the ontology to the<br /> grid fault diagnosis, combining the traditional method,<br /> they established a diagnostic system based on the<br /> ontology failure event model. The system is able to reveal<br /> the deep reason in power plant failures.<br /> The D. D. Dou et al. [31] proposed a fault diagnosis<br /> method based on ontology, explained the ontology-based<br /> fault diagnosis process, presented ontology-based fault<br /> knowledge representation and fault diagnosis algorithm.<br /> In the instance, the failure ontology consisted of 1000<br /> conceptions.<br /> In an automobile fault diagnosis expert system, N. Y.<br /> Shi et al. [32] used the ontology to realize knowledge<br /> sharing. The investigations showed that the different<br /> knowledge representation would cause the heterogeneity<br /> in various systems then be separated from each other and<br /> can not share and reuse the knowledge. This increased the<br /> difficulty of system development. So they also tried to<br /> use ontology to solve the heterogeneity of the expert<br /> system to realize the knowledge sharing.<br /> L. Du et al. [33] applied ontology to construct an<br /> automotive engine fault diagnosis expert system.<br /> According to the reasoning characteristics of automotive<br /> engine fault diagnosis system, they discussed the<br /> construction of the domain ontology knowledge library,<br /> structure definition and system reasoning description.<br /> Q. Niu et al. [34] proposed a knowledge description<br /> method basing on description logic in motor fault<br /> diagnosis, and the logical error detection reasoning has<br /> been done to the motor fault knowledge library. It can<br /> efficiently represent the relationship of the motor fault<br /> diagnosis domain knowledge and check out the error of<br /> knowledge logic system. During the experiment, the a<br /> diagnosis system was developed using the ontology<br /> editing tools protégé and OWL language, at the same<br /> <br /> Journal of Automation and Control Engineering, Vol. 1, No. 2, June 2013<br /> <br /> time, realized the logical error detection reasoning<br /> through TABLEAU algorithm.<br /> There are so many applications of ontology in fault<br /> diagnosis such as chemical process [35], AC motor [36],<br /> gear-hobbing machine [37], the air compressor [38]-[39],<br /> other rotating machinery [40], and so on. And it has<br /> captured the attentions of researchers in the field of<br /> pumping system [41].<br /> In our study, an ontology model for sucker-rod<br /> pumping system running state fault diagnosis system was<br /> proposed. The knowledge management system<br /> framework is built up, which includes data layer, logic<br /> layer and application layer. The comprehensive diagnosis<br /> ontology database, including the knowledge domains of<br /> fault, collection and structure, was established. And the<br /> relationships among the ontology were described.<br /> III.<br /> <br /> reasoning for ontology library and consistency detecting.<br /> The application layer is based on the logical layer, and<br /> the application program is intent to provide services of<br /> fault diagnosis information query, statistics and decision<br /> support system for ordinary users. The manager can edit<br /> the structure and content of the ontology through a<br /> specific interface.<br /> IV.<br /> <br /> PUMPING SYSTEM FAULT KNOWLEDGE<br /> REPRESENTATION [38]<br /> <br /> According to the characteristics of the pumping system<br /> fault diagnosis elements, it can be divided into three<br /> knowledge domains: fault domain (to express the<br /> knowledge of pumping system operating status fault<br /> classification, fault causes, and so on), collecting<br /> information domain (to express the acquisition<br /> information and operational knowledge of the pumping<br /> system) and structure domain (to express the knowledge<br /> about the hardware configuration of the pumping system).<br /> According to the characteristics of the structure and<br /> content of the information, the different information<br /> sources can be realized knowledge representation in its<br /> own ontology respectively. The heterogeneous<br /> knowledge of the three domain ontology is interrelated<br /> but all are basing on the fault integrated diagnosis<br /> ontology knowledge. The components of the domain<br /> ontology are shown in Fig. 2.<br /> <br /> FAILURE KNOWLEDGE MANAGEMENT SYSTEM<br /> FRAMEWORK<br /> <br /> In view of present oilfield production and organization<br /> structure characteristic, basing on ontology, this paper<br /> puts forward a pumping system running state fault<br /> diagnosis knowledge management system framework is<br /> show in Fig. 1.<br /> <br /> Figure 2. Domain ontology information constitutes<br /> <br /> Figure 1. Knowledge management system framework<br /> <br /> From Fig. 1 we can see that this model consists of a<br /> data layer, logic layer and an application layer. The<br /> relational database and integrated diagnostic ontology<br /> one in data layer are used to store related information of<br /> pumping system fault diagnosis.<br /> The ontology database includes fault domain ontology,<br /> collection information domain ontology and structure<br /> domain ontology and they are three heterogeneous<br /> domain ontology. The ontology data adapter is for data<br /> exchange between ontology databases and data sets,<br /> while providing operation method for the interaction of<br /> the data in the ontology database. The logical layer data<br /> mapping platform is used to build the mapping relation<br /> between the traditional relational database and ontology<br /> library. The interface of ontology management and user<br /> can provide methods of access and modify ontology<br /> database for users. According to the data from the<br /> management interface, the ontology generation module<br /> generates a specified ontology structure. The inference<br /> engine is then capable of the heterogeneous ontology<br /> <br /> A. Fault Integrated Diagnosis Ontology<br /> The fault integrated diagnosis ontology (FIDOnto) is a<br /> knowledge management model in which using ontology<br /> model to represent all kinds of information and their<br /> relationship with relevant to the fault diagnosis. It is<br /> expressed as:<br /> FIDOnto::=(Id,C, A, R, S, Ax)<br /> where Id identifies the name of the domain ontology, C is<br /> the concept set, A is the attribute set, R is the relation set,<br /> S is the body node status set and Ax is the axioms set of<br /> the field.<br /> FIDOnto describes the common characteristics of all<br /> objects in a field. It is the parent ontology of the local<br /> ontology within this field, being equivalent to a<br /> knowledge base class. The fault domain ontology,<br /> collecting information domain ontology and structure<br /> domain ontology inherit the properties of the parent<br /> ontology, and been added to new members to describe the<br /> faults discipline from different views.<br /> <br /> 128<br /> <br /> Journal of Automation and Control Engineering, Vol. 1, No. 2, June 2013<br /> <br /> B. Fault Domain Ontology<br /> The fault domain ontology (FaultOnto) is a knowledge<br /> management to describe the recorded information about<br /> the operational status failure mode of pumping systems,<br /> causes or workover. It is expressed as:<br /> <br /> makeUseOf, causedBy, independentOf, testAt. K<br /> indicates the connection credibility (0